Object detection result determination method and object detection model training method and apparatus
By performing feature extraction, candidate box prediction, and iterative information interaction processing on multiple panoramic images, the problem of insufficient utilization of multi-view information in existing technologies is solved, achieving more efficient and accurate 3D detection.
CN115171062BActive Publication Date: 2026-07-24BEIJING PHIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING PHIGENT TECHNOLOGY CO LTD
- Filing Date
- 2022-06-01
- Publication Date
- 2026-07-24
AI Technical Summary
Technical Problem
Existing 3D detection methods lack the utilization of multi-view information, resulting in poor detection performance. At the same time, processing each view individually leads to low detection efficiency.
Method used
By inputting multiple panoramic images into the object detection model, feature extraction, candidate box prediction, and candidate box optimization layers are used to perform feature extraction, candidate box prediction, and iterative information interaction processing, thereby achieving the fusion and optimization of multi-view information.
Benefits of technology
It improves the efficiency and performance of 3D detection, enabling parallel processing of multi-view image information and enhancing the accuracy and speed of detection results.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115171062B_ABST
Abstract
The application provides an object detection result determination method and an object detection model training method and device. The method comprises: inputting a plurality of surround view images of a target vehicle collected to an object detection model; the object detection model comprises: a feature extraction layer, a candidate box prediction layer and a candidate box optimization layer; calling the feature extraction layer to perform feature extraction processing on the plurality of surround view images to obtain feature maps of each surround view image at different scales; calling the candidate box detection layer to perform candidate box prediction processing on the plurality of feature maps at different scales to obtain predicted candidate box positions and predicted candidate box features corresponding to the plurality of surround view images; and calling the candidate box optimization layer to perform information iteration interaction processing on the predicted candidate box positions and the predicted candidate box features corresponding to the plurality of surround view images to obtain an object detection result of the surroundings of the target vehicle. The application can improve the performance and efficiency of three-dimensional detection.
Need to check novelty before this filing date? Find Prior Art